Guanchu (Gary) Wang
|
Ph.D. Candidate |
 |
About me
I am currently a PhD. candidate in Rice University supervised by Professor Prof. Xia Hu. Before joining Rice University, I was working as a research assistant in Texas A&M University (Prof. Xia Hu group) in 2020 and Westlake university (Professor Donglin Wang group) in 2019. From 2012 to 2019, I recevied B.S. degree from Dalian University of Technology (DUT) and M.S. degree from University of Science and Technology of China (USTC), where I was fortunate for having been supervised by Prof. Chen Gong and Zhenyuan Xu.
My primary research focus is in the realm of Trustworthy AI, a critical area that demands the infusion of trust throughout the AI lifecycle. Within this overarching theme, my research focuses on explainable AI (XAI), developing algorithms to faithfully explain the behaviors of deep neural networks. My research can be widely applied to most foundation models, including multilayer perception, graph neural networks, convolutional neural networks, vision transformers, and large language models.
In addition to my primary focus, I collaborate closely with health informaticians from UTHealth and Baylor College of Medicine, leveraging AI to address critical challenges in healthcare.
I will be on the job market in Fall 2024 seeking for academic opportunities.
Education
- From 2021.9 to now: Ph.D. candidate, Computer Science, Rice University.
- From 2016.9 to 2019.6: M.S., Information Science and Technology, University of Science and Technology of China (USTC).
- From 2012.9 to 2016.6: B.S., Information and Communication Engineering, Dalian University of Technology (DUT).
Research Experience
- From 2024.5 to 2024.8: Research intern, VISA Research, Foster City, CA, USA.
- From 2020.9 to 2021.8: Graduate research assistant, Texas A&M University, USA.
- From 2019.9 to 2020.1: Research assistant, Westlake University, Hangzhou, P.R.C.
What's New
- 10/2024: Our paper Taylor Unswift: Secured Weight Release for Large Language Models via Taylor Expansion. has attracted 175.9K views and 1.7K like on Twitter; 1K like on Xiaohongshu; and 5.6K impressions and 70 reactions on LinkedIn🔥! See [Twitter] [Xiaohongshu] [LinkedIn]
- 09/2024: Our work Taylor Unswift: Secured Weight Release for Large Language Models via Taylor Expansion. is accepted by EMNLP 2024! Credits to all co-authors! See [Paper].
- 09/2024: Our benchmark paper KV Cache Compression, But What Must We Give in Return? A Comprehensive Benchmark of Long Context Capable Approaches. is accepted by EMNLP 2024 Findings! Credits to all co-authors! See [Paper].
- 08/2024: Our open-source dataset, corpora, and benchmark paper on LLMs for rare disease diagnosis have been released! See [Dataset] [Corpora] [Code] [Paper].
- 07/2024: Our work Efficient GNN Explanation via Learning Removal-based Attribution is published on TKDD! See [Paper] and [Code].
- 07/2024: Our benchmark paper on KV Cache Compression has been released! See [Paper] and [Code].
- 06/2024: Our benchmark paper on training foundational time-series models has been released! See [Paper], [Code], [Twitter], [Medium], and [Zhihu].
- 05/2024: Our work TVE: Learning Meta-attribution for Transferable Vision Explainer is accepted by ICML 2024! See [Paper].
- 03/2024: My pleasure of receiving the Doctoral Forum Travel Award in SDM 2024 (Society for Industrial and Applied Mathematics)!
- 02/2024: It is my honor to have received the admission of Future Leader Summit 2024!
- 11/2023: Our Natural Language UI for the NORD Rare-disease Database is online! [APP].
- 10/2023: Our work DiscoverPath: A Knowledge Refinement and Retrieval System for Interdisciplinarity on Biomedical Research achieves the Best Paper Honorable Mention in CIKM 2023! Thanks to all of the co-authors! See [Paper].
- 09/2023: Our work Winner-Take-All Column Row Sampling for Memory Efficient Adaptation of Language Model is accepted by NeurIPS 2023! See [Paper].
- 09/2023: Our work Chasing Fairness under Distribution Shift: a Model Weight Perturbation Approach is accepted by NeurIPS 2023! See [Paper].
- 08/2023: Our work DiscoverPath: A Knowledge Refinement and Retrieval System for Interdisciplinarity on Biomedical Research is accepted by CIKM 2023 demo track! See [Paper] [Code] [Demo] [FrontEnd].
- 06/2023: Our work Mitigating Algorithmic Bias with Limited Annotations is accepted by ECML-PKDD 2023! See [Paper].
- 04/2023: Our work DIVISION: Memory Efficient Training via Dual Activation Precision is accepted by ICML 2023! See [Paper].
- 03/2023:Invited poster in the Energy High Performance Computing Conference sponsored by the Ken Kennedy Institute.
- 03/2023:Invited talk in the seminor of CS department at Rice University. See [Talk].
- 02/2023: Our efficient XAI survey paper has been reported by TechBeat and Tencent Developer! See [TechBeat] and [Tencent].
- 02/2023: Our opensource project DiscoverPath has been released! See [DiscoverPath] for online demo; and [Github] for source code.
- 01/2023: Our work CoRTX: Contrastive Learning for Real-time Explanations. is accepted by ICLR 2023! See [Paper]
- 12/2022: It's my honor to have received the KKI 2022/23 Graduate Fellowship from the Ken Kennedy Institute of Rice University. See [the Full List].
- 10/2022: Our real-time object detection demo work achieves the Best Paper Award in CIKM 2022! Thanks to all of the co-authors! See [Paper].
- 09/2022: Our Accelerating Shapley Explanation via Contributive Cooperator Selection is selected as the Spotlight of ICML 2022! See [Paper].
- 08/2022: Tutorial of medical anomaly detection using TODS has been released! See [Github].
- 08/2022: Our real-time object detection system based on MAX78000 is accepted by CIKM 2022 Demo Track! See [Github].
- 06/2022: Our work Accelerating Shapley Explanation via Contributive Cooperator Selection is accepted by ICML 2022! See [Paper].
- 12/2021: Our real-time object detection based on MAX78000 for edge devices has been released! See [Github].
- 10/2021: Two papers Fairness via Representation Neutralization and Revisiting Time Series Outlier Detection: Definitions and Benchmarks are accepted by NeurIPS 2021!
- 09/2020: Our opensource package TODS is accepted by AAAI 2021 Demo Track! See [Paper].
- 06/2020: Opensource package TODS for end-to-end time-series outlier detection has been online! See [Github].
- 04/2020: Our work Independent Skill Transfer for Deep Reinforcement Learning is accepted by IJCAI 2020 (Accepted rate 12.6%)! See [Paper].